Feasibility of automated treatment planning for HyperArc-based stereotactic recurrent head and neck cancers using knowledge-based planning, multicriteria optimization, and dosimetric scorecards

The development of a HyperArc-based automated stereotactic treatment planning process for recurrent head and neck (RHN) cancers using knowledge-based planning, multicriteria optimization (MCO), and dosimetric scorecards. Twenty-eight previously treated RHN HyperArc patients, prescribed 30 to 40 Gy in 5 fractions, were used to generate a RapidPlan model. An Eclipse Scripting script was created to automate the model’s use and refine the model’s output with MCO, where tradeoff exploration was guided by a dosimetric scorecard. Final dose, calculated with AcurosXB, was normalized to the original plan’s PTV D95 %. Maximum doses to organs-at-risk (OARs), target metrics, calculation/optimization times and patient-specific quality assurance results were recorded. The automated treatment planning process took on average 12.1 ± 0.9 minutes to generate the RHN RapidPlan plan and an additional 25.0 ± 2.6 minutes for MCO exploration. The GTV mean dose stayed comparable across all plans with an average decrease of 0.8 ± 0.9 Gy from the original plan to the RapidPlan and an average increase of 0.3 ± 1.2 Gy from the original plan to the MCO plan. The PTV Paddick’s conformity index received an average decrease of 0.01 ± 0.01 from original to RapidPlan and an increase of 0.02 ± 0.01 from the original plan to the MCO plan. All OAR maximum dose changes were favorable or remained within clinically acceptable dose levels for stereotactic delivery. The dose delivery accuracy for the final plans were verified with electronic portal imaging device-based portal dosimetry and an independent Monte Carlo second check reported averages of 97.2 % and 96.3 %, respectively, indicating accurate stereotactic delivery. The process established throughout this work offers an automated treatment planning routine, excluding contouring, for stereotactic RHN patients.